Deep Learning-Based Auction Algorithms and the Applications to Data Science Platforms

Abstract

This paper introduces auction algorithms those are actively used for distributed resource allocation, and then discusses the applications to data science platforms. Among various auction algorithms, second price auction (SPA) is known to be truthful, thus it is generally used. However, this second price auction is with one drawback where the auctioneer cannot obtain maximum revenue. Thus, deep learning based approximation for the SPA can be used for various resource allocation problems, and this paper discusses the applications to data science platforms.